1. Plot vector geospatial data (points/polygons)
2. Plot raster geospatial data (image)
3. Extracting data from raster
4. Making maps (base maps, compass rose, overlays, etc)

We are researchers who want to answer the question "What range of environmental conditions do zebra tend to inhabit?"

General workflow:

1. Plot locations of zebra
2. Plot map of environmental conditions
3. Plot contextual information
4. Extract data about environmental conditions at those locations

Variants of this question/workflow are applicable to other study systems.

• Read and write spatial data
• Represent geographic and attribute data
• Transform between different models of Earth
• Do geographic operations
• Make maps

• Vector data model uses points, lines, and polygons (e.g. locations of animals, ecosystem types) and can include nonspatial attribute data
  • How many counts at a point?
  • Length of a transect
  • Land-cover type of a polygon
• Raster data model divides surface into cells of constant size (e.g. gridded temperature)

• Many tools for spatial analysis in R
• sf package for vector data
  • Developed in 2016 with support from R consortium
  • Based on simple features ISO standards that are not R specific
  • Replaces sp, rdgal for read/write, and rgeos for geometric operations
  • Integrates well with tidyverse
• raster package for raster operations and more
  • Released in 2010
  • Can support processing very large files

• sf (simple feature) objects are extended data.frames or tibbles
  • attribute data stored as tibble
  • geometry stored as a list column
  • can have multiple types of geometry in one object
• class is sfc (simple feature columns) with useful components like
  • bounding box bbox
  • CRS (coordinate reference system) attributes

• sf functions begin with st_
• Methods for summary, plot
• sf methods for filter, arrange, distinct, group_by, ungroup, mutate

methods(class = "sf")
##   [1] $<-                        [                         
##   [3] [[<-                       aggregate                 
##   [5] anti_join                  arrange                   
##   [7] as.data.frame              cbind                     
##   [9] coerce                     dbDataType                
##  [11] dbWriteTable               df_spatial                
##  [13] distinct                   extent                    
##  [15] extract                    filter                    
##  [17] full_join                  gather                    
##  [19] group_by                   group_map                 
##  [21] group_split                identify                  
##  [23] initialize                 inner_join                
##  [25] left_join                  mask                      
##  [27] merge                      mutate                    
##  [29] nest                       plot                      
##  [31] print                      raster                    
##  [33] rasterize                  rbind                     
##  [35] rename                     right_join                
##  [37] sample_frac                sample_n                  
##  [39] select                     semi_join                 
##  [41] separate                   separate_rows             
##  [43] show                       slice                     
##  [45] slotsFromS3                spread                    
##  [47] st_agr                     st_agr<-                  
##  [49] st_area                    st_as_sf                  
##  [51] st_bbox                    st_boundary               
##  [53] st_buffer                  st_cast                   
##  [55] st_centroid                st_collection_extract     
##  [57] st_convex_hull             st_coordinates            
##  [59] st_crop                    st_crs                    
##  [61] st_crs<-                   st_difference             
##  [63] st_force_polygon_cw        st_geometry               
##  [65] st_geometry<-              st_interpolate_aw         
##  [67] st_intersection            st_intersects             
##  [69] st_is                      st_is_polygon_cw          
##  [71] st_join                    st_line_merge             
##  [73] st_linesubstring           st_make_valid             
##  [75] st_minimum_bounding_circle st_nearest_points         
##  [77] st_node                    st_normalize              
##  [79] st_point_on_surface        st_polygonize             
##  [81] st_precision               st_segmentize             
##  [83] st_set_precision           st_simplify               
##  [85] st_snap                    st_snap_to_grid           
##  [87] st_split                   st_subdivide              
##  [89] st_sym_difference          st_transform              
##  [91] st_transform_proj          st_triangulate            
##  [93] st_union                   st_voronoi                
##  [95] st_wrap_dateline           st_write                  
##  [97] st_zm                      summarise                 
##  [99] transmute                  ungroup                   
## [101] unite                      unnest                    
## see '?methods' for accessing help and source code

roads <- st_read(here("data", "enproads.shp"))
## Reading layer `enproads' from data source `C:\Users\mbuon\Desktop\QERM597-making-maps\data\enproads.shp' using driver `ESRI Shapefile'
## Simple feature collection with 565 features and 3 fields
## geometry type:  LINESTRING
## dimension:      XY
## bbox:           xmin: 14.35337 ymin: -19.48418 xmax: 17.15124 ymax: -18.5008
## epsg (SRID):    NA
## proj4string:    NA
class(roads)
## [1] "sf"         "data.frame"
head(roads)
## Simple feature collection with 6 features and 3 fields
## geometry type:  LINESTRING
## dimension:      XY
## bbox:           xmin: 14.46841 ymin: -19.41752 xmax: 14.58964 ymax: -19.33077
## epsg (SRID):    NA
## proj4string:    NA
##     TYPE LENGTH KM                       geometry
## 1  Track    0.7  1 LINESTRING (14.50502 -19.35...
## 2  Track    3.6  4 LINESTRING (14.49077 -19.36...
## 3 Graded    2.3  2 LINESTRING (14.4864 -19.417...
## 4  Track    7.1  7 LINESTRING (14.58498 -19.34...
## 5  Track    1.2  1 LINESTRING (14.58498 -19.34...
## 6 Graded    0.1  0 LINESTRING (14.52596 -19.33...


# Highlight parts
# 1. list column (aka sfc)
# 2. feature geometry (sfg)
# 3. feature (row)

# Default methods for objects
plot(roads)

plot(filter(roads[,3], KM > 20))

• Packages such as rnaturalearth
• Convert from other Spatial* objects using st_as_sf

world <- ne_countries(scale = "medium", returnclass = "sf")
st_geometry(world) %>% plot()

Default data is Lat/Long

Namibia <- ne_countries(country = "namibia")
Namibia_sf <- st_as_sf(Namibia)

plot(Namibia_sf)

* Exercise: try for another country

#Uncomment here if using rnatural earth for the first time:
# devtools::install_github("ropenscilabs/rnaturalearthhires", force = TRUE)
# install.packages("rnaturalearthhires", repos = "http://packages.ropensci.org")

namibia_political <- ne_states(country = "namibia")
namibia_political_sf <- st_as_sf(namibia_political)
plot(namibia_political_sf)

• st_crs() used to convert coordinates (do projections) and do datum transformations
• Proj.4 string or ESPG code contains this info
• ESPG code: public registry of spatial reference systems
• Web repositories for ESPG and Proj.4 information
• ESPG codes identify coordinate reference systems (CRS)
  • Ex: CRS= 3857 (web-based mapping i.e. Google, OpenStreetMap)
• Lat/Long data:
  • Often expressed as WGS84: World Geodic System 1984
  • Ex: CRS = 4326 (coordinate system based on Earth's center of mass)
• ESPG = European Petroleum Survey Group

st_crs("+proj=longlat +datum=WGS84") # Proj.4 string"
## Coordinate Reference System:
##   EPSG: 4326 
##   proj4string: "+proj=longlat +datum=WGS84 +no_defs"
st_crs(3857)                         # ESPG = 3857
## Coordinate Reference System:
##   EPSG: 3857 
##   proj4string: "+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgrids=@null +wktext +no_defs"
st_crs(3857)$units                   # check units
## [1] "m"
st_crs(4326)
## Coordinate Reference System:
##   EPSG: 4326 
##   proj4string: "+proj=longlat +datum=WGS84 +no_defs"
st_crs(4326)$units
## NULL
st_crs(NA)                          # unknown (assumed planar/Cartesian)
## Coordinate Reference System: NA

• st_transform transforms or converts coordinates to new reference system

(a1 <- st_area(Namibia_sf))
## 824886560992 [m^2]
(a2 <- st_area(st_transform(Namibia_sf, 32733))) # Namibia plane m (UTM)
## 826175760745 [m^2]
(a3 <- st_area(st_transform(Namibia_sf, 4293))) # Another reference from espg.io 
## 824892086801 [m^2]
(a4 <- st_area(st_transform(Namibia_sf, 4326)))
## 824886560992 [m^2]
# Namibia projection = 4326

*If you are working with U.S. data you can convert to U.S. feet

Change the units!

units::set_units(a1, km^2)
## 824886.6 [km^2]
units::set_units(a1, ft^2)
## 8.879005e+12 [ft^2]

zebra_data <- read.csv(here("data", "zebra.csv"))
head(zebra_data)
##   X.1   X  Name            Date Longitude  Latitude Speed Direction Temp
## 1  50  78 AG059 4/20/2009 15:40  15.89605 -19.12138     0         0    0
## 2  34  60 AG059 4/20/2009 13:01  15.91854 -19.17761     0         0    0
## 3  33  59 AG059 4/20/2009 12:02  15.91857 -19.17766     0         0    0
## 4  62  90 AG059 4/20/2009 17:00  15.80543 -19.08002     0         0    0
## 5 109 159 AG059  4/21/2009 1:40  15.91887 -19.17758     0       270    0
## 6 107 155 AG059  4/21/2009 1:00  15.91882 -19.17760     0       270    0
##   Altitude PDOP    ID         DatePOS diff       day
## 1     1123    2 AG059 4/20/2009 15:40   19 4/20/2009
## 2     1113    3 AG059 4/20/2009 13:01   59 4/20/2009
## 3     1130    3 AG059 4/20/2009 12:02  721 4/20/2009
## 4     1139    2 AG059 4/20/2009 17:00   17 4/20/2009
## 5     1133    2 AG059  4/21/2009 1:40   40 4/21/2009
## 6     1126    4 AG059  4/21/2009 1:00   20 4/21/2009

zebra_sf <- read.csv(here("data", "Zebra.csv")) %>% 
  dplyr::select(ID = Name, 4:6) %>% 
  mutate(timestamp = as.POSIXct(lubridate::mdy_hm(Date))) %>%
  st_as_sf(., coords = 3:4, crs = "+init=epsg:4326") 

st_transform(zebra_sf, 32733) #convert to UTM
## Simple feature collection with 11490 features and 3 fields
## geometry type:  POINT
## dimension:      XY
## bbox:           xmin: 546746.9 ymin: 7852220 xmax: 691330.5 ymax: 7912185
## epsg (SRID):    32733
## proj4string:    +proj=utm +zone=33 +south +datum=WGS84 +units=m +no_defs
## First 10 features:
##       ID            Date           timestamp                 geometry
## 1  AG059 4/20/2009 15:40 2009-04-20 15:40:00 POINT (594243.6 7885501)
## 2  AG059 4/20/2009 13:01 2009-04-20 13:01:00   POINT (596576 7879266)
## 3  AG059 4/20/2009 12:02 2009-04-20 12:02:00 POINT (596579.5 7879260)
## 4  AG059 4/20/2009 17:00 2009-04-20 17:00:00   POINT (584733 7890124)
## 5  AG059  4/21/2009 1:40 2009-04-21 01:40:00 POINT (596611.3 7879269)
## 6  AG059  4/21/2009 1:00 2009-04-21 01:00:00 POINT (596605.8 7879266)
## 7  AG059  4/21/2009 5:01 2009-04-21 05:01:00 POINT (596604.9 7879271)
## 8  AG059  4/21/2009 4:41 2009-04-21 04:41:00 POINT (596610.9 7879268)
## 9  AG059  4/22/2009 4:00 2009-04-22 04:00:00 POINT (596603.6 7879269)
## 10 AG059  4/22/2009 2:01 2009-04-22 02:01:00 POINT (596609.4 7879272)

countries <- ne_countries(scale=110)

p <- ggplot() +  geom_polygon(data=countries, aes(x=long, y=lat, group=group),  color="white", lwd = .25)
 
ggplotly(p)

• shapefiles
• database connections
• use st_write

1. Load the roads shapefile (enproads.shp) and upload Zebra.csv and plot the geometry.
2. If we look at our roads data, we see that there are several types of roads in the shapefile. How close or far from different road types do zebra move?

roads <- st_read(here("data", "enproads.shp"), crs = "+init=epsg:4326") %>% #4326= coordinate system based on Earth's center of mass
  st_transform(32733) #32733 = spatial reference for Nambia 
## Reading layer `enproads' from data source `C:\Users\mbuon\Desktop\QERM597-making-maps\data\enproads.shp' using driver `ESRI Shapefile'
## Simple feature collection with 565 features and 3 fields
## geometry type:  LINESTRING
## dimension:      XY
## bbox:           xmin: 14.35337 ymin: -19.48418 xmax: 17.15124 ymax: -18.5008
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs

st_geometry(roads) %>% plot

zebra_sf <- read.csv(here("data", "Zebra.csv")) %>% 
  dplyr::select(ID = Name, 4:6) %>% 
  mutate(timestamp = as.POSIXct(lubridate::mdy_hm(Date))) %>%
  st_as_sf(., coords = 3:4, crs = "+init=epsg:4326") %>% #longlat, converting foreign object to SF object
  st_transform(32733) #convert to UTM

ggplot() +
  geom_sf(data=roads) +
  geom_sf(data=zebra_sf, aes(color=ID))

How close or far from different road types do zebra move?

unique(roads$TYPE)
## [1] Track  Graded Gravel Tar   
## Levels: Graded Gravel Tar Track

#Filter roads based off type:
large_roads <- filter(roads, TYPE %in% c("Tar", "Gravel"))
small_roads <- filter(roads, TYPE %in% c("Graded", "Track"))

ggplot() +
  geom_sf(data=large_roads, size=1.5) + 
  geom_sf(data=small_roads, size=0.6) + 
  geom_sf(data=zebra_sf, aes(color=ID))

# Find the minimum distance (in meters) of each point to a large road
large_dist<- st_distance(y=zebra_sf, x=large_roads) # a units matrix dim =[nrow(x), nrow(y)]; takes 10-20 seconds
zebra_sf$large_road_dist <- apply(large_dist, 2, min)

# Find the minimum distance (in meters) of each point to a small road
small_dist<- st_distance(y=zebra_sf, x=small_roads) # a units matrix dim =[nrow(x), nrow(y)]; takes 10-20 seconds
zebra_sf$small_road_dist <- apply(small_dist, 2, min)

head(data.frame(zebra_sf))
##      ID            Date           timestamp                 geometry
## 1 AG059 4/20/2009 15:40 2009-04-20 15:40:00 POINT (594243.6 7885501)
## 2 AG059 4/20/2009 13:01 2009-04-20 13:01:00   POINT (596576 7879266)
## 3 AG059 4/20/2009 12:02 2009-04-20 12:02:00 POINT (596579.5 7879260)
## 4 AG059 4/20/2009 17:00 2009-04-20 17:00:00   POINT (584733 7890124)
## 5 AG059  4/21/2009 1:40 2009-04-21 01:40:00 POINT (596611.3 7879269)
## 6 AG059  4/21/2009 1:00 2009-04-21 01:00:00 POINT (596605.8 7879266)
##   large_road_dist small_road_dist
## 1        7.845860       3479.8987
## 2      163.187201        241.8660
## 3      156.746071        245.8446
## 4        9.115608       1605.2154
## 5      151.769751        227.9759
## 6      151.358102        231.6984

ggplot(zebra_sf) +
  geom_histogram(aes(large_road_dist, fill="large roads"), alpha=0.5) +
  geom_histogram(aes(small_road_dist, fill="small roads"), alpha=0.5) +
  scale_fill_manual(labels=c("large roads", "small roads"), values=c("blue", "orange"))

Assume you have an existing 'matrix' object in your workspace that you want to cast as a 'raster' object.

wspd.raster<-raster(wspd,xmn=x.min,ymn=y.min,xmx=x.max,ymx=y.max)

class(wspd)
## [1] "matrix"
class(wspd.raster)
## [1] "RasterLayer"
## attr(,"package")
## [1] "raster"

plot(wspd.raster)

You should have a file labeled "ndvi_mean_utm.tif" in your working directory. Use ?raster and upload the raster into your workspace. Plot a simple raster map.

Solution:

NDVI_mean <- raster(here("data", "ndvi_mean_utm.tif"))
NDVI_mean
## class      : RasterLayer 
## dimensions : 474, 1274, 603876  (nrow, ncol, ncell)
## resolution : 231.6564, 231.6564  (x, y)
## extent     : 431873.8, 727004, 7844489, 7954294  (xmin, xmax, ymin, ymax)
## crs        : +proj=utm +zone=33 +south +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0 
## source     : C:/Users/mbuon/Desktop/QERM597-making-maps/data/ndvi_mean_utm.tif 
## names      : ndvi_mean_utm 
## values     : -558.117, 5303.021  (min, max)

Solution:

plot(NDVI_mean)

ggplot(data.frame(rasterToPoints(wspd.raster))) +
  geom_raster(aes(x=x, y=y,fill=layer))

Take the raster object and plot it in ggplot. Can you overlay the zebra location data from earlier? (EC) Can you write equivalent code using the pipe (i.e. %>%) functionality?

Solution: (no piping)

ggplot(data.frame(rasterToPoints(NDVI_mean))) +
  geom_raster(aes(x=x, y=y,fill=ndvi_mean_utm)) +
  geom_sf(data=zebra_sf, size=0.7,col="black")

Solution: (piping)

NDVI_mean %>% rasterToPoints() %>% data.frame()  %>% ggplot()

Solution: (no piping)

ggplot(data.frame(rasterToPoints(NDVI_mean)))

pop_sf <- read_csv(here("data", "huc_whp_burnable_area_population.csv")) %>% 
  st_as_sf(., coords = 6:5, crs = "+init=epsg:4326") 
## Parsed with column specification:
## cols(
##   huc_name = col_character(),
##   whp = col_double(),
##   acres_burnable = col_double(),
##   population_in_huc = col_double(),
##   Latitude = col_double(),
##   Longitude = col_double()
## )
pop_sf$wspd <- raster::extract(wspd.raster, as(pop_sf, "Spatial"))

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   2.130   2.927   3.361   3.442   3.974   5.414

Modify this code to extract NDVI information at the Zebra locations. Calculate summary statistics.

Solution:

zebra_sf$ndvi<-raster::extract(NDVI_mean, as(zebra_sf, "Spatial"))
summary(zebra_sf$ndvi)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   626.5  2527.6  2682.2  2775.8  3015.1  4212.1

slice<-raster(ncol=20,nrow=20)
slice[]<-rnorm(n=ncell(slice))
sim_stack <- stack(slice,slice^2,1/slice)

dates<-c(1:3)
#assign time value (in this case year) to each layer
sim_stack <- setZ(sim_stack, dates)

#time match and extract
stack_dates <- getZ(sim_stack) #all raster layer dates
zebra_sf$NDVI <- NA
for (i in 1:50) { #this would take a long time to run through nrow(zebra_sf) locations, so demonstrating for first 50 locations
  print(paste0("extracting zebra location #",i))
  zeb_date <- as.Date(zebra_sf$timestamp[i]) 
  stack_idx <- which(stack_dates %in% seq(zeb_date - 15, zeb_date, by='day')) ##which raster layer date matches zebra date
  zebra_sf$NDVI[i] <-raster::extract(NDVI_stack[[stack_idx]], #as(zebra_sf$geometry[i], "Spatial")) #extract
}
head(data.frame(zebra_sf))

Using netCDF files (ncdf4 package in R) are my preferred approach.

Focus on ggmap

• As of mid-2018, requires registering with Google and obtaining an API key
• Requires providing a valid credit card (yikes!), though charges are nonexistent or at least very minimal
• See the ggmap GitHub page for more information about API keys

• Basemaps available from Google, Stamen, or Open Street Map
• Terrain, satellite, or watercolor
• See this helpful cheat sheet

Two steps:

• Download basemap raster
• Plot raster and overlay other spatial data

This is my personal API key, which you can use for the purposes of this class!

register_google(key = "AIzaSyBRkw_wzDKuWZDikdD86Kmp1Sa6PzuCKFc")

To add a Stamen basemap, first define the bounding box for the basemap you would like to download.

zebra_bbox <- c(14, -20, 17.5, -18)
names(zebra_bbox) <- c("left", "bottom", "right", "top")

terr_basemap <- get_stamenmap(bbox=zebra_bbox, zoom=9, maptype="terrain")

ggmap(terr_basemap) +
  geom_sf(data=roads, inherit.aes = FALSE) + 
  geom_sf(data=zebra_sf, aes(color=ID), inherit.aes = FALSE) +
  coord_sf(crs = st_crs(4326))

wat_basemap <- get_stamenmap(bbox=zebra_bbox, zoom=9, maptype="watercolor")

ggmap(wat_basemap) +
  geom_sf(data=roads, inherit.aes = FALSE) + 
  geom_sf(data=zebra_sf, aes(color=ID), inherit.aes = FALSE) +
  coord_sf(crs = st_crs(4326))

To add a Google basemap, first define the center coordinates for the basemap you would like to download.

zebra_center <- c(15.8, -19)
names(zebra_center) <- c("lon", "lat")

sat_basemap <- get_googlemap(center=zebra_center, zoom=8, size=c(640, 640), scale=2,
                             maptype="satellite")
ggmap(sat_basemap) +
  geom_sf(data=roads, inherit.aes = FALSE) + 
  geom_sf(data=zebra_sf, aes(color=ID), inherit.aes = FALSE) +
  coord_sf(crs = st_crs(4326))

Notice that get_googlemap() returns a square basemap tile, which then sets the plotting limits of your map.

To manually set different plotting limits, pass in an empty "base layer" and set x and y limits using ggplot().

ggmap(sat_basemap, maprange=FALSE, base_layer=ggplot(aes(x=1, y=1), data=NULL)) +
  xlim(14.3, 17.4) + ylim(-20, -18) + xlab("lon") + ylab("lat") +
  geom_sf(data=roads, inherit.aes = FALSE) + 
  geom_sf(data=zebra_sf, aes(color=ID), inherit.aes = FALSE) +
  coord_sf(crs = st_crs(4326))

• One alternative package is RgoogleMaps
  • No need to get Google API key (yay!)
  • Does not seems to be compatible with ggplot2 (boo)
• Another approach is to download raster data directly from Natural Earth
  • Have to download manually from website and import tif file as a raster brick, then display red/green/blue values
  • Only works well if you are mapping a large spatial extent (lacks fine resolution)

1. Adjust the zoom level in any of the previous code examples
   • zoom = integer from 3-21
   • 3 = continent, 10 = city, 21 = building
2. Try using the geocode function to find the lat/lon coordinates for a known location
   • e.g., geocode("Space Needle")

my_basemap <- get_googlemap(center="Space Needle", zoom=14, size=c(640, 640),
                            scale=2, maptype="satellite")

ggmap(my_basemap)

• Legends for spatial layers
• Add a scale bar and north arrow
• Inset maps

q0 <- ggplot() +
  geom_sf(data = large_roads, size = 1.5, color = "grey", inherit.aes = FALSE) + 
  geom_sf(data = small_roads, size = 0.6, color = "grey", inherit.aes = FALSE) + 
  geom_sf(data = zebra_sf, aes(color = ID), inherit.aes = FALSE, 
          alpha = 0.1,               # fix over plotting
          show.legend = "point")     # make legend keys points instead of rects.

q1 <- q0 +
  coord_sf(xlim = c(14, 17.5),   # make space for legend
           ylim = c(-20.5, -18.25), 
           expand = FALSE,
           crs = st_crs(4326)) + 
  scale_color_manual(
    values = viridis::viridis(10), # pick new color scale
    name = "Zebra ID",             # change title of legend
    guide = guide_legend(
      direction = "horizontal", # horizontal legend
      title.position = "top",   # title at top
      title.hjust = 0.5,        # center title
      byrow = TRUE,             # organize legend keys
      nrow = 3,                 # organize legend keys
      override.aes = list(alpha = 1) # make the legend keys 100% opaque
    )) +
  theme_bw() +
  theme(legend.position = c(0.65, 0.2),        # position legend in lower R corner
        legend.key = element_rect(fill = NA), # remove background of legend keys
        panel.grid.major = element_line(color = "white")) + # remove graticules
  labs(x = "Longitude", y = "Lattitude")

q2 <- q1 +
  annotation_scale(location = "bl") +  # from ggspatial package
  annotation_north_arrow(location = "bl", which_north = "true", 
                         pad_x = unit(0.75, "in"), pad_y = unit(0.5, "in"),
                         style = north_arrow_fancy_orienteering)

ne_namibia <- ne_countries(scale = "medium", country = "namibia", returnclass = "sf")

n1 <- ggplot(ne_namibia) + # need to put data in ggplot call for geom_rect
  geom_sf() +
  geom_sf(data = large_roads, inherit.aes = FALSE) +
  geom_sf(data = small_roads, inherit.aes = FALSE) +
  geom_rect(xmin = 14, xmax = 18, ymin = -20, ymax = -18, # extent indicator
            fill = NA, colour = "black", size = 1.5) +
  theme_bw() +
    coord_sf(xlim = c(11, 47),       
           ylim = c(-34, -16.9), # make additional space for inset map
           expand = FALSE,
           datum = NA)           # remove graticules and coords

# Calculate ratio of inset plot
# use st_bbox(zebra_sf)
inset_ratio <- (-18.5 - (-20.5))/(17.5 - 14)

myfig <- cowplot::ggdraw(n1) +
  cowplot::draw_plot(q2, width = 1.2, height = 1.2*inset_ratio,
            x = 0.01, y = 0.1, hjust = 0, vjust = 0)

• Simple Features package
  • Vignettes, blogs, presentations
  • Cheatsheet
  • issue tracker
• r-spatial Edzer Pebesma's blog
• Spatial Data Science
• Geocomputation with R
  • excellent in-depth treatment
  • includes theory, examples, and covers rasters